Solutions
Find the right product for your situation.
Two distinct buyer problems, two products. Each stands alone — you don't need one to use the other.
GPU inference fleets
You run GPU inference at scale.
Cold starts are costing you latency SLOs. Power caps are limiting how aggressively you can prewarm. You need a scheduler that manages both at once.
Typical operators
Hyperscalers, GPU clouds, frontier labs, AI factories running vLLM / TensorRT-LLM / SGLang at scale.
What changes
Demand is forecast before it hits. Warmth is managed as inventory. Preparation is scheduled against your watt cap — not just "spin up more replicas."
Entry path
Pilot with telemetry and demand forecasting first. No automation until your team enables it.
Honesty
Helm is in pilot and projection phase. Fleet automation is off by default. We measure carefully before we claim.
Hardware leasing & finance
You finance or lease AI hardware.
Every asset. Live covenant status. An audit trail that holds up to scrutiny.
Typical buyers
Equipment lessors, captive finance, OEM finance, institutional lenders, and internal audit teams for organizations owning large GPU estates.
What changes
Live covenant compliance status per asset. A Blake3-linked audit trail for every state change. No more spreadsheets before a quarterly review.
Entry path
Docker pilot with sample data (under an hour). DCGM push for live fleet data typically the same day with your team. No Aurora inference required.
Honesty
Passport is standalone. You do not need Helm or any Aurora inference product to use it.
Not sure which fits?
Tell us about your infrastructure and we'll route you to the right starting point.